Will single-read be enough coverage?

[Turnerac0987]

I apologize if this question has already been asked, but I couldn't find the exact info I was looking for, so I decided to just ask my specific question directly.

I'm trying to set up a run on a GAIIx with 16 samples using 2 lanes, so 8 samples in each lane. We're doing targeted sequencing (not whole exome) and there are a total of 922,000 amplicon bases in the complete design (560,000 unique loci bases). We're trying to get 30X coverage, 20X coverage at a minimum. We're doing a run length of 72bp and we're thinking of doing single read.

What I'm trying to figure out:

Is single read going to provide enough coverage of these samples or should we do a paired-end run? (We're sharing a flowcell with another group and they're doing single read)

How do I figure out how many reads I need to get 30X coverage of 8 samples per lane?

I hope I'm asking my questions right and providing the right info. This is the first time I'm figuring this out by myself so I just want to make sure I'm loading enough of each sample, and whether I'm going to get enough coverage on single read.

Thanks so much for your help!

[Simon Anders]

Why do you think paired-end would get you better coverage than single-end? You still see the same number of fragments, only now, you see both ends of each fragment. Whether this extra information is of any value for you depends on what you intend to do with your data.

[swbarnes2]

The sample prep is the same amuont of time and money whether you do paired and or single. Paired end doubles the time spent on the instrument (so it goes from about 3 days to about 7), and doubles the number of reagents used, and doubles your data.

I usually work out the calcualtions this way: You have about a Mb of sequence. If your reads are, say, 60 long, and you had a million of them covering a million bases of reference, that'd be about 60x coverage. You can get 30-40 million clusters at least, on a single lane. And if you do paired end, that's two reads per cluster.

But also keep in mind that if you sequence a million reads, they will not all fall on target. It's probably depend on the target probes, and your lab, but figure about 30-60% of the reads will land on target. So I guess for 30x coverage, you want about about a million reads per sample.

[cedance]

Why do you think paired-end would get you better coverage than single-end? You still see the same number of fragments, only now, you see both ends of each fragment. Whether this extra information is of any value for you depends on what you intend to do with your data.

Simon, a question. Fragments that contain exon-exon junctions, under sufficient insert size, the pairs should map back to both these exons separately, right? If we calculate coverage by the number of reads covered for each base, then wouldn't we be counting both these reads separately, and consequently more coverage (per exon)?

[Simon Anders]

Yes, but if you had twice as many single-end reads you would get even more coverage. And spreading one sample over two lanes and sequencing single-end costs about the same as doing one paired-end lane, I think. Hence, if you don't want to acively use the information which exon got paired with which, it still does not seem clear to me that PE is more economical.

[cedance]

I agree. I guess, I'd consider it economical or not depending on the type of task at hand. If I am looking at splicing events, or isoform abundance, then ideally I would like to get rid of all the pcr duplicates. I guess paired end reads mapping to the same position with same or exact insert size tells me clearer that they are highly likely to be pcr duplicates than many single end reads mapping at the same position. I don't know if this is of concern to all possible studies possible to RNA-Seq though.

[Turnerac0987]

Thanks for all the help guys! I have a much better understanding of how to calculate coverage and number of reads. Looks like single read will be more than enough for aligning to a reference to find SNPs.